How to make classes form numerical data for deep learning training
Show older comments
I want to make classes from my target data, the input and target data are numerical, I know that I can make 3 intervals but what should the format of the classes be ? In the matlab’s examples the classes are categorical format, should I make them also categorical in my case ? Thanks
6 Comments
Wabi Demeke
on 3 Aug 2021
Edited: Wabi Demeke
on 3 Aug 2021
Interval_1= 0, Interval_2=1, and Interval_3=2
Output=[1 0 2 1...2]' somthing like this then
Outputs=categorical(Output); % use to convert to categorical
I think it should be ready to train using classificationLayer
Omailnoth
on 3 Aug 2021
Wabi Demeke
on 3 Aug 2021
Edited: Wabi Demeke
on 3 Aug 2021
I thought you wanted to label each outputs data to 3 classes based on their intervals for example ouputs range between 30-50 to class_1(=0), 51-70 to class_2(=1), and 71-90 class_3(=2). And then to train NN that predict which class a certain input belongs.
if you want to train NN that predict the actual ouputs(could be any value between 30-90) instead of which ranges it belongs you can just use regression NN.
Omailnoth
on 3 Aug 2021
Wabi Demeke
on 3 Aug 2021
Excatly! Follow common NN training procedures such as spliting your dataset into training and validation set to check your NN generlization capacity. You probably need to spend sometime in hyperparametric tuning to find NN model with good prediction performance. Goodluck.
Peter Perkins
on 3 Aug 2021
Not sure what the question is here, but to make a categorical from numeric, discretize is the best way to do it.
Answers (1)
Shivang Srivastava
on 11 Aug 2021
0 votes
As per my understanding you want to convert discreet / numerical data to intervals for Training Deep Learning Models.
Categories
Find more on Deep Learning Toolbox in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!